A vehicle-road cooperation intelligent scheduling optimization method

Abstract

This invention discloses an intelligent scheduling optimization method for vehicle-road cooperation, relating to the field of intelligent traffic control. The method determines whether a target road segment is a backlit section based on information from vehicle dashcams and collects sunlight or road surface reflection information for the corresponding backlit section. Before a vehicle enters the backlit section, a machine learning-based glare compensation time prediction model is invoked to estimate the glare compensation time based on the sunlight or road surface reflection information and vehicle information. The vehicle's current speed is multiplied by the glare compensation time to obtain the potential danger distance. Vehicles in the vehicle group whose distance to the vehicle in front is less than the potential danger distance are identified and included in a potential risk vehicle set. The speed of this potential risk vehicle set is adjusted to ensure its potential danger distance is greater than the distance to the vehicle in front. This invention can predict the driver's visual delay risk before a vehicle enters a backlit area, enabling early intervention and reducing the risk of accidents caused by glare.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control, and more specifically, to an intelligent scheduling optimization method for vehicle-road cooperation. Background Technology

[0002] Under certain lighting conditions, especially when a vehicle enters a backlit area while driving, the driver's visual reaction can be severely affected. Backlighting is usually caused by direct sunlight or strong reflections from the road surface. When the sun's azimuth angle is close to the road's heading angle, a bright area appears in front of the vehicle, causing the driver to experience momentary blindness or blurred vision. At this time, the driver's visual system needs time to adapt to the change in light. If the driver fails to perceive the dynamic changes of the vehicle ahead within this time, delayed braking or a lag in reaction may occur.

[0003] In real-world traffic scenarios, especially when multiple vehicles are queuing, the reaction delays of individual drivers can propagate through the vehicle group, causing cascading decelerations and even rear-end collisions. Existing intelligent driving assistance systems typically rely solely on individual vehicle sensors to detect the distance to the vehicle ahead, failing to consider differences in driver visual responses caused by lighting conditions and lacking group risk control mechanisms for backlit scenarios. Furthermore, the intensity and reflectivity of backlit areas vary with time and location, making it difficult for a single static model to accurately describe the actual risks. How to assess driver reaction delays based on real-time lighting information, and then perform dynamic risk identification and control at the vehicle group scale, has become a key challenge in current vehicle-road cooperative safety control. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent scheduling optimization method for vehicle-road cooperation, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A vehicle-road cooperative intelligent scheduling optimization method includes the following steps:

[0007] Based on the information from the dashcams of the vehicle group, determine whether the target road section is a backlit section, and collect the sunlight information or road surface reflected light information of the corresponding backlit road section;

[0008] Before a vehicle enters the backlit road section in a group of vehicles, a glare compensation time prediction model based on machine learning is invoked to estimate the glare compensation time when the vehicle enters the backlit road section based on the sunlight information or road surface reflected light information of the backlit road section and vehicle information.

[0009] Multiply the vehicle's current speed by the glare compensation time to obtain the potential danger distance;

[0010] Identify vehicles in the vehicle group whose distance from the vehicle in front is less than the potential danger distance and include them in the potential risk vehicle set. Adjust the speed of the potential risk vehicle set so that its potential danger distance is greater than the distance from the vehicle in front.

[0011] In some embodiments, the dashcam in each vehicle in the vehicle group continuously collects forward-facing video.

[0012] When the video image features of any vehicle are analyzed and determined to be a backlight scene, the corresponding backlight information is recorded. The backlight information includes the vehicle's geographical location, driving direction and timestamp, as well as the corresponding sunlight information and road surface reflected light information collected by the dashcam.

[0013] The vehicle broadcasts the backlight information to the same group of vehicles;

[0014] After receiving backlight information, the vehicle compares its own position with its driving direction. When it is expected to enter a backlight section within a preset time window, it will determine in advance that it is about to enter a backlight section and trigger the glare compensation time prediction.

[0015] In some embodiments, the sunlight information includes solar azimuth angle, solar altitude angle, and light intensity.

[0016] In some embodiments, the road surface reflected light information includes the ratio of road surface reflected brightness to background brightness.

[0017] In some embodiments, the determination of the backlit road section includes the following steps:

[0018] If, in a series of consecutive frames, the rate of change of the exposure time of the dashcam for sunlight or road surface reflected light is negative and exceeds the absolute value of the set threshold, while the rate of change of the photosensitivity or brightness compensation parameter is positive and exceeds the absolute value of the set threshold, then the camera device is determined to enter the high brightness suppression state.

[0019] When the duration of the high brightness suppression state exceeds the preset time window, the corresponding road segment is confirmed to be a backlight road segment.

[0020] In some embodiments, the training data for the glare compensation time prediction model is derived from vehicle group operation data;

[0021] Each vehicle in the vehicle group collects driving videos through its vehicle-mounted driving recorder. Based on the video image features, backlight scenes are identified, and the time interval from when the driver starts to press the brake to when the driver releases the brake is extracted as a single glare response sample, combined with the brake pedal signal when the vehicle enters a backlight scene.

[0022] Samples from multiple vehicles are collected to form a label set for training in supervised learning.

[0023] In some embodiments, the glare compensation time prediction model is trained using machine learning based on training samples. The input features include sunlight information or road surface reflected light information and vehicle information corresponding to each sample, and the output features include the predicted value of glare compensation time.

[0024] In some embodiments, the vehicle information includes any one or more of the following: vehicle model, windshield transmittance, and driver's seat viewpoint height.

[0025] In some embodiments, the remaining vehicles in the vehicle group outside the potentially risky vehicle group are classified based on the difference between the distance to the vehicle in front and the potential danger distance, and divided into a low-risk group, a medium-risk group, and a high-risk group, wherein the larger the difference, the lower the risk.

[0026] The speed control range is ordered from largest to smallest as follows: the high-risk set is larger than the medium-risk set, which is larger than the low-risk set.

[0027] The advantage of this invention over existing technologies lies in its proposed intelligent scheduling optimization method for vehicle-road cooperation. This method enables dynamic risk assessment and hierarchical control to address the driver's visual reaction delay when a vehicle enters a backlit road section. This mechanism allows the system to predict risks before a vehicle enters a backlit area and issue control commands in advance, effectively preventing reaction delays and rear-end collisions caused by glare, thus improving overall road safety.

[0028] This invention utilizes forward-facing video image information from vehicle-mounted dashcams to identify backlighting scenarios caused by direct sunlight or strong road surface reflections in real time during natural vehicle driving, and automatically labels the spatial location and temporal characteristics of backlighting areas. Compared to methods relying on fixed roadside sensors or weather models, using dashcams offers advantages such as wide coverage, high real-time performance, low cost, and dynamic updates. The system can cross-validate backlighting determination results through data from multiple vehicles, thereby forming a high-confidence backlighting area mapping, providing accurate data for subsequent glare compensation time prediction and risk control. This mechanism enables vehicle groups to identify and share backlighting information in advance, achieving early risk warning and collaborative scheduling, thereby effectively reducing reaction delays and rear-end collisions caused by glare, and significantly improving overall safety and traffic stability in a vehicle-road cooperative environment.

[0029] Another significant advantage of this invention lies in the training method of the glare compensation time prediction model. The system learns from dashcam video and vehicle braking data during vehicle convoy operation. By automatically identifying backlighting scenarios and extracting the time interval from when the driver presses and releases the brake as a sample, it can collect a large amount of effective data in real-world driving environments without repeated experiments. This method not only saves data acquisition costs but also significantly improves the model's generalization ability and scene adaptability.

[0030] Furthermore, this invention introduces a method of dividing vehicle groups into different risk levels based on potential danger distances and implementing tiered speed control according to the risk level. High-risk vehicles are given priority to decelerate, while medium- and low-risk vehicles maintain a steady following distance, thereby ensuring overall traffic flow while maintaining safe distances. Compared to traditional single-vehicle control strategies, this invention can achieve group-level risk coordination in dynamic environments with multiple vehicles cooperating, and has high practical value and promotional significance. Attached Figure Description

[0031] Figure 1 This is an overall structural diagram of the invention;

[0032] Figure 2 This is a structural diagram of the backlight recognition and broadcasting of the present invention;

[0033] Figure 3 This is a diagram of the glare compensation time prediction and training structure of the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0035] This invention addresses the backlight safety problem in vehicle-road cooperative environments by proposing a complete technical approach consisting of identification, prediction, evaluation, and control.

[0036] like Figure 1 The diagram shown illustrates the specific implementation of the invention method, including:

[0037] Based on the information from the dashcams of the vehicle group, determine whether the target road section is a backlit section, and collect the sunlight information or road surface reflected light information of the corresponding backlit road section;

[0038] Before a vehicle enters the backlit road section in a group of vehicles, a glare compensation time prediction model based on machine learning is invoked to estimate the glare compensation time when the vehicle enters the backlit road section based on the sunlight information or road surface reflected light information of the backlit road section and vehicle information.

[0039] Multiply the vehicle's current speed by the glare compensation time to obtain the potential danger distance;

[0040] Identify vehicles in the vehicle group whose distance from the vehicle in front is less than the potential danger distance and include them in the potential risk vehicle set. Adjust the speed of the potential risk vehicle set so that its potential danger distance is greater than the distance from the vehicle in front.

[0041] In a more specific embodiment, the dashcams of the vehicle group are used to continuously acquire forward video on the target road section, and the presence of backlighting is determined by image analysis and the temporal characteristics of the camera device parameters.

[0042] In some embodiments, sunlight and road surface reflected light information can be extracted directly from the geometric and brightness information of the dashcam video. Specifically, the vehicle's driving direction, i.e., the dashcam's shooting direction, can be obtained from the vehicle's GPS and electronic compass, while the sun's direction is determined by the centroid location of the highlighted area in the image. The angle between the vehicle's direction and the sun's direction represents the sun's azimuth angle. For example, when this angle is close to 0° and the vehicle is traveling west, the sun's azimuth angle is due west. The sun's altitude is obtained based on the angle between the sun's direction recorded in the dashcam and the vehicle's horizontal direction. Illumination intensity can be measured by the dashcam's automatic metering system. In engineering implementation, dashcams typically have a photoelectric sensor in front of the image acquisition chip or use the pixel signals of the image sensor itself to calculate the average brightness value. The camera internally calculates a metering value based on the overall brightness of the image, which is used for automatic exposure adjustment. If the metering value continuously increases and the automatic exposure time decreases accordingly, it indicates that the external light is increasing; if the metering value decreases and the exposure time increases, it indicates that the light is decreasing. In another embodiment, a photosensitive element can be directly installed inside the vehicle or on the roadside to measure the light intensity; the road surface reflected light information is obtained by analyzing the ratio of the light intensity of the ground area to the light intensity of the background area. The background area generally refers to the area above the road surface, such as the sky above the road surface; the higher the ratio, the higher the degree of glare.

[0043] In a further embodiment, the automatic determination of backlit road sections relies on the stable characteristics of consecutive frames rather than instantaneous peaks. The system constructs a time series for three quantities: exposure time, ISO, and brightness compensation. If, within several consecutive frames, the rate of change of exposure time is negative and its absolute value exceeds a threshold, while the rate of change of ISO or brightness compensation is positive and its absolute value exceeds a threshold, and this high-brightness suppression state continues for more than a preset time window, the corresponding location can be confirmed as a backlit road section. The reason for this setting is that the dashcam automatically enters a high-brightness suppression state in backlit environments to prevent overexposure. When a vehicle enters a backlit area from a normally lit area, the camera detects direct or reflected strong light and quickly shortens the exposure time to reduce the amount of light entering the camera, while simultaneously increasing ISO or brightness compensation to maintain image detail. Therefore, the exposure time shows a significant decreasing trend, while ISO or brightness compensation shows an increasing trend, and this change is continuous in the time series. By jointly judging the dynamic changes of these three quantities, it is possible to effectively distinguish between short-term brightness fluctuations and true backlighting scenes. Only when the high brightness suppression state lasts for a certain time window is it identified as a backlighting section, thereby improving the stability and reliability of the judgment.

[0044] The threshold and time window can be set according to the camera frame rate and the degree of ambient light fluctuation. For example, if the inter-frame change exceeds 10% and lasts for more than half a second, the misjudgment can be significantly reduced.

[0045] like Figure 2As shown, after the judgment is completed, the system records the backlight information, including the geographical location, driving direction, and timestamp of the triggering vehicle, as well as the sunlight information and road surface reflection information corresponding to the segment, and broadcasts it to the same vehicle group through the vehicle-to-infrastructure (V2I) network. Subsequent vehicles, after receiving the information, combine it with their own location and driving direction to make estimations for the navigation system. If it is expected that they will enter the backlight segment within a preset time window, they enter the prediction preparation stage.

[0046] like Figure 3 As shown, during the prediction preparation phase, the system invokes a machine learning-based glare compensation time prediction model before the vehicle reaches the backlight boundary. The model inputs are sunlight information or road surface reflected light information corresponding to the backlight section, and vehicle information, including vehicle model, windshield transmittance, and driver's seat viewpoint height. Using these parameters, the model can generate the glare compensation time at the moment of entry, characterizing the time required for the driver's vision to recover from the glare's impact to reliable recognition.

[0047] In a specific embodiment, the model is trained from the long-term accumulation of vehicle group operation data. The method involves automatically labeling backlight segments in dashcam videos and simultaneously reading the brake pedal signals when the vehicle enters a backlight scene. The time interval from pressing the brake to releasing the brake is extracted as a single glare response sample. By aggregating samples from multiple vehicles and multiple scenes, a label set is formed. Under a supervised learning framework, the mapping between input features and glare compensation time is learned.

[0048] A glare compensation time prediction model can be implemented using a multilayer perceptron neural network. This model consists of an input layer, several hidden layers, and an output layer. The input layer receives three feature sets: sunlight information, road surface reflected light information, and vehicle information. The hidden layers employ a fully connected structure and use a non-linear activation function to enhance the model's ability to fit complex lighting and vehicle characteristics. The output layer is a single-node structure, corresponding to the predicted glare compensation time.

[0049] After obtaining the glare compensation time, the system calculates the potential danger distance by multiplying the vehicle's current speed by that time. The potential danger distance represents the blind travel distance the vehicle may generate at its current speed during the visual recovery period, and is used to compare it with the actual distance to the vehicle in front. If the actual distance is less than the potential danger distance, the system determines that the vehicle poses a rear-end collision risk upon entering the backlighting zone and includes it in the potential risk vehicle set. For vehicles in this set, the cooperative terminal or the vehicle terminal itself issues a speed control command to limit the vehicle's speed and control the vehicle so that the potential danger distance is not less than the distance to the vehicle in front, thus avoiding potential rear-end collisions after entering the backlighting area.

[0050] In a further embodiment, the present invention classifies the remaining vehicles in the vehicle group outside the potential risk vehicle set based on the difference between the distance to the vehicle in front and the potential danger distance, and divides them into a low-risk set, a medium-risk set, and a high-risk set, wherein the larger the difference, the lower the risk; wherein the speed control range is ordered from large to small as follows: the high-risk set is greater than the medium-risk set, which is greater than the low-risk set.

[0051] In some embodiments, the risk level is classified according to the difference between the distance to the vehicle in front and the potential danger distance. For example, a distance greater than 10 meters is classified as low risk, a distance between 5 and 10 meters is classified as medium risk, and a distance less than 5 meters is classified as high risk.

[0052] In specific embodiments, high-risk group vehicles decelerate the most, and the system can issue commands through the vehicle-road cooperative control module or the vehicle terminal itself. In some embodiments, the deceleration can be reduced within the range of 3% to 8% of the speed. Medium-risk group vehicles are slightly larger than the potential danger distance, and deceleration is smaller. In some embodiments, the deceleration can be between 2% and 4%. Low-risk group vehicles have sufficient distance, and only slight speed adjustments or maintaining the current speed are required, with the adjustment range controlled within 1%.

[0053] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent scheduling optimization of vehicle-road cooperation, characterized in that, The method comprises the following steps: determining whether the target section belongs to a glare section based on the information of the vehicle group's driving recorders and collecting the sunlight information or road surface reflection light information of the corresponding glare section; before the vehicle enters the glare section, calling a machine learning-based glare compensation time prediction model to estimate the glare compensation time when the vehicle enters the glare section based on the sunlight information or road surface reflection light information of the glare section and the vehicle information; multiplying the current speed of the vehicle by the glare compensation time to obtain a potential danger distance; finding vehicles in the vehicle group that have a distance to the preceding vehicle less than the potential danger distance and including them in a potential risk vehicle set, and adjusting the speed of the potential risk vehicle set so that the potential danger distance is greater than the distance to the preceding vehicle; the training data of the glare compensation time prediction model is derived from the vehicle group operation data; each vehicle in the vehicle group collects driving video through the vehicle-mounted driving recorder, identifies the glare scene based on the video image features, and extracts the time interval from when the driver starts to step on the brake to when the brake is released as a single glare reaction sample in combination with the brake pedal signal of the vehicle when entering the glare scene; collecting samples of multiple vehicles to form a label set for supervised learning training; the glare compensation time prediction model is trained based on the training samples using machine learning, the input features include the sunlight information or road surface reflection light information and vehicle information corresponding to each sample, and the output features include the predicted value of the glare compensation time.

2. The method of claim 1, wherein, the driving recorder of each vehicle in the vehicle group continuously collects forward video; when the video image features of any vehicle are analyzed and determined to be a glare scene, the corresponding glare information is recorded, including the geographical position, driving direction and time stamp of the vehicle, and the corresponding sunlight information and road surface reflection light information collected by the driving recorder; the vehicle broadcasts the glare information to the same vehicle group; after receiving the glare information, the subsequent vehicle compares its own position and driving direction, and when it is predicted to enter the glare section within a preset time window, it is determined to be about to enter the glare section and triggers the glare compensation time prediction.

3. The method according to claim 1 or 2, characterized in that, The sunlight information includes the sun azimuth angle, the sun elevation angle and the illumination intensity.

4. The method according to claim 1 or 2, characterized in that, The road surface reflection light information includes the ratio of the road surface reflection brightness to the background brightness.

5. The method of claim 4, wherein, The determination of the glare section comprises the following steps: in a plurality of consecutive image frames, if the change rate of the exposure time of the driving recorder with respect to the sunlight or the road surface reflection light is negative and exceeds the absolute value threshold, and the change rate of the photosensitivity or brightness compensation parameter is positive and its absolute value exceeds the set threshold, it is determined that the camera enters a high-light suppression state; when the duration of the high-light suppression state exceeds the preset time window, the corresponding section is confirmed as a glare section.

6. The method of claim 1, wherein, The vehicle information includes any one or more of the following: vehicle model, front windshield light transmittance, driver's eye point height.

7. The method of claim 1, wherein, The method further comprises: classifying the remaining vehicles in the vehicle group outside the potential risk vehicle set based on the difference between the distance to the preceding vehicle and the potential danger distance, and dividing them into a low-risk set, a medium-risk set and a high-risk set, wherein the larger the difference, the lower the risk.

8. The method of claim 7, wherein, The vehicle group is hierarchically regulated, wherein the amplitude of speed regulation is sorted as follows from large to small: the high-risk set is greater than the medium-risk set, which is greater than the low-risk set.

Images